A Multi-Agent System Approach for Predicting Investment Decisions in Land Use Formation and Building Typologies

Abstract

Urban land use formation and building typologies emerge from complex interactions between planning regulations, household investment behavior, and market risks. Yet existing models often treat these processes in isolation, limiting their predictive capacity for post–land readjustment development. Here we present a unified simulation framework that integrates Cellular Automata (CA), Portfolio Theory with Bayesian probability, and Multi-Agent Systems (MAS) to model how landowners select land uses and building forms under varying risk–return scenarios. Treating residential plots as the fundamental decision unit, the model reproduces theoretical patterns of portfolio choice—capturing high-risk/high-return and low-risk/low-return outcomes—while maintaining stability across parameter variations. Applied to a district in Kanazawa City as an empirical validation, the framework demonstrates strong predictive accuracy in simulating household investment strategies and resulting urban configurations. Beyond case-specific results, this research establishes a transferable methodology that links land use dynamics with investment decision-making, offering planners and policymakers a tool to explore future urban scenarios and guide sustainable development.